This was stupid as it prevented the compiler from doing code analysis
and discovering the bug leading to a segmentation fault when using
compiler optimization (see christian/shellipt#44).
Long overdue update of `README.md`.
I think that might be the culprit for the segmentation fault in Nix
(see christian/shellipt#44). Chances are good, because the when I fixed
the problem, I bypassed those faulty macros.
Laying out the windows in the window manager frame, is a very similar
process regardless of doing it horizontally or vertically. So the
template design pattern was applied and the general functionality was
pulled up; only specific functionalities were put into the corresponding
classes.
kNCursesDemoApp: Separated the screen into three windows vertically. The
top and bottom window contain another two windows each, lay out
horizontally. The top left window is fixed to width 1 and the top right
takes the rest of the space. Bottom two windows take 50 % of the space
each.
Remarks: There are still some rendering bugs, when hiding and showing
windows. E. g. when hiding the top or middle window with `<F1>` or
`<F2>` respectively, the bottom window has some errors. Maybe some
timing issue.
`TilingWindowManager` was created from `VerticalTilingWindowManager` and
separated its functionality so that there can be a
`HorizontalTilingWindowManager` next.
* switching to recursive mutexes for now; easier to implement
* need to read user input from the window manager; hidden windows are
drawn when reading input from there
* note: occasional dead lock between PTY and ncurses mutex
* fixed type in class `SingleUserInput`
* not depending on "-dev" packages for the built Debian package
* some refactoring in if-clauses and with "quotes"
* libvterm dependency: if something is missing, made clear what
Some other changes to the CMake-files done as well:
* documentation looks more like Doxygen-style
* prevent mix-up with C++ GSL library ("Guidelines support library" not
"GNU scientific library")
There is no version detection. If there is any vterm library, it is
used. If it is the wrong version, there will be an build error.
`libvterm.so.0.0.2` is preferred, which hopefully is compatible to
revision 740 or Ubuntu `libvterm0` version 0.1.2-2.
If no library is found, the correct one is downloaded and build.
* using `get_wch()` now and mapping most of the function keys documented
in `vterm_keycodes.h` (not num keypad)
* there are still a lot of mappings from ncurses missing (in conjunction
with modifier keys, related issue #11)
* using mutex to make sure writing to terminal client is serialized
* Output strings with unprintable characters either as a table or as
continuing string of characters; the byte value of unprintable
characters (`isprint(<character>, <locale>) == false`) is shown.
* Output readable string of `key` -> `value` pairs with an enumerated
type for the key and union type for the value.
When pressing return in `nano` the terminal received an `<CTRL>+J` instead of
the return key. `<CTRL>+J` justifies the text. But is (confusingly) actually
the key for a line feed as well? Terminal programming is not easy …
Christian Burger